Smart and sustainable logistics for perishable goods: Integrating deep learning, blockchain, and TOPSIS in coffee supply chains

Efficient and sustainable logistics for perishable goods in remote and uncertain environments presents complex challenges that require advanced technological integration. This study introduces a smart and sustainable logistics framework for managing specialty coffee supply chains in Northern Thailand, combining deep learning, blockchain, and multi-criteria decision analysis. A hybrid optimization approach is developed to address a dynamic, multi-objective vehicle routing problem (VRP) under uncertainty. An autoencoder is used to generate high-quality initial routes, while a Generative Adversarial Network (GAN) adapts them in real-time to handle disruptions such as traffic and weather variability. TOPSIS is applied to evaluate trade-offs among key sustainability objectives, including travel time, carbon emissions, and product freshness. Blockchain infrastructure ensures secure, transparent coordination across stakeholders, while smart contracts and IoT-based tracking support real-time decision-making and traceability. Using real-world data from 200 coffee shops, 10 supply hubs, and 15 delivery vehicles over a 30-day period, the proposed system achieved a 21.42% reduction in travel time, a 35.87% cut in emissions, and a 78.56% improvement in customer satisfaction. This research offers novel insights into how emerging technologies can enhance sustainability in last-mile logistics for perishable goods.